TimeFinder / SNAP Questions

1. Multiple VDEVS can share a pool of SAVE devices. When you create the VDEV pairings you define which SAV Pool you want to use, so if the pool only has a single SAVE device in it I guess you would be sharing that one.

2. The SAVE device should be sized to accommodate the rate of change on the source device for the SNAP. If your data is expected to change by 30% during the SNAP session you may want to size the SAVE device at 40% of the size of the source device (to allow a bit of leeway)

3. The SAVE device doesn't read data. The host will read data through the VDEV. The VDEV maintains pointers to the original data. If the original data changes the original will get written to the SAVE device and the pointer for that data will be updated on the VDEV.

4. I believe you can just make a VDEV of the same size (without the configuration being identical), but I'm not absolutely certain right now. I'll have a look and see if I can find something definitive.

5. No. If you issue the Terminate the SNAP session will cease to exist and you won't be able to read the data. It isn't like a Clone or BCV where you actually create a copy of the data. The SNAP is just pointer based, so you can't terminate until you are done with the SNAP.

Hope this helps.

Basically yes.

You should have a look at the Solutions Enabler Timefinder Product Guide for the full details, but you're on the right track with the basics.

Just make sure you size the SAVE pool large enough to accommodate the rate of change. Since there is no real limit on the size of SAV devices within the pool (just on the total capacity of the pool to meet your needs) you can always see if you have some "spare" space lying around that is unused on the physical drives with your current configuration and make some smaller SAV devices (smaller than the STD and BCV devices you normally use. Someone correct me if I'm wrong on this, but this could well be an efficient way to get some use out of "wasted space" on the physical drives.

Thanks! Just to clarify for question # 5. I should create the SNAP and Activate it, then leave it alone while I read the data. When we want to resync up the data again I should unmount the drive, terminate, then active again. Is that right?

Hi Not / Hi Allen,

Just my 2-cents worth. The VDEV's can be activated against devices of equal or smaller size. The SAV devices should be protected and Raid-5 is fine (but 2-MIR protection may yield better overall SNAP performance). They can be ANY convenient size (below Enginuity 5772 they are limited to the same size or smaller than the largest VDEV). Yes, any convenient free space can be used for SAV devices (a common practice ). But bear in mind that SAV pool data is striped across all SAV pool members so if you "mix" a number of different sizes in the same pool you may see variations in performance as the pool "fills up".

Regards,
Michael.

my 1.5 euro-cents



You have to create 8 VDEVS (8 GB each) and form them in a meta having the same config (concat/striped) of the STD volume you want to snap.


-s-

Message was edited by:
Stefano Del Corno

Do you have any documentation to back this up?

I'm not trying to start argument, I just want to understand where you are getting this from. If it is true I want to be able to plan for this in the future.

BTW- Right now I do create all my VDEVs absolutely identical to the STDs they are snapping. It just works out that way because of the way we pre-create the devices and VDEVs. It is easier to build them together with the same config.

Hi Allen,

My apologies for causing confusion. I always worry about writing too much and creating confusion. Unfortunately too short a posting can have the same effect

As you and Stefano have pointed out the need for identical VDEV meta volumes to be identically constructed is in the Solutions Enabler TimeFinder Product Guide. No problem there.

My comments on using larger VDEV's being legal comes from our EMC internal Symmetrix configuration training material:

CDEVs/ VDEV's can be activated against devices of equal or
SMALLER size.
Virtual device can be established to a RAID-10 head as long
as it is larger or equal to the size of the entire RAID-10 volume.

But I has not tried testing this statement first hand and assumed it was also in the TimeFinder Product Guide.

However, my statement (and our internal training documentation) is incorrect.

I just tried a "symsnap create" with a 18414 cylinder VDEV and 4602 cylinder STD and received:

"The size of the Source device is not equal to the size of the Target device".

I don't know if this changes at Solutions Enabler 6.4.2 (which now specifically allows Clone targets to be larger):

Clone source and target devices no longer have the requirement to be
the same size. Now the size of a clone target device can be larger than
the source device. This support requires the following SYMCLI
environment variable be set:
SYMCLI_CLONE_LARGER_TGT = ENABLED

But I would have to ask someone else to test this out... At this point please disregard my earlier (incorrect) comments on larger VDEV's being allowed by the API.

Thanks & Regards,
Michael.

Hi Michael and everybody else...



Shame on you .. Internal docs should remain internal



AFAIK this allows you to use bigger hyper .. but the geometry of the meta MUST still be the same. So you can snap a 64 Gb meta (8 concat members, 8Gb each) with a 72 Gb vdev-meta (8 concat members, 9 Gb each), just like you can have R2 larger then R1 devices. I'm still wondering how the hosts will use the exceeding capacity .. But it's not my problem right now



Your post wasn't inaccurate.. was a flash-forward from the future .. and requires everybody to put a lot of care in explaining this brand new feature

No problem Michael,

It happens! As far as you were concerned you were propagating valid info. That's why I always find it valuable when people post references to the source of info when there is any question. In this case you had no reason to believe your source was not entirely accurate (at least not in ALL situations).

Thanks Allen,

I appreciate that - I should know what I am talking about and I just hate the idea of giving out bad advice (of course, you can't test everything personally and need to take some information at "face" value).

I did speak with Mainframe Software Support and they stated that Mainframe TimeFinder SNAP does not care that the target can be larger than the source (virtual or otherwise since they don't differentiate). The mainframe SNAP VOLUME command has a REFVTOC option specifically to cater for target volumes that are physically larger than the source.

Back to Open Systems. At SE 6.4.2 (with the appropriate environment variable set) you can certainly Clone from smaller source to larger target BUT I would wait to try this on a VDEV before considering changing your "regular" VDEV size. The syscalls used by the SYMAPI to check for the legality of a symclone "create" amd symsnap "create" are different.

Thanks again.

Regards,
Michael.